Multi-camera Video Synopsis In Geographic Scene

With the popularity of smart cities,surveillance video plays an increasingly important role in people’s lives.Video synopsis methods,which are used to ease the conflict of the large amount of video data with sparsely distributed valuable information,are also in constant innovation.There are some disadvantages with traditional video synopsis methods based on image space,such as limited camera field of view,pixel contention in the process of image expression,and backward technology.Along with the development of technology,there are higher requirements for monitoring scope and geographical environment characteristics,which also leads to the development of video synopsis technology become more professional(3D geographic scene),wide(wider monitoring range)and special(following the trend of computer and geographic information technology).Due to the discontinuity of video objects’ trajectories between multi-camera(which caused by the blind area between the cameras’ field of view),trajectories clustering algorithms based on multi-camera video need to be improved;Because of the independence of video data and geographical scene between cameras,the global retrieval method of video objects remains to be completed;What’s more,as topological relationship between cameras’ location and cameras’ field of view need to be explored,the multi-camera video synopsis methods are expected to become more perfected.In the background of above problems,this paper mainly launches research from four following aspects:1.Multi-camera video data information can be extracted through computer vision technology which could make cross camera video data associated with each other.Geographic information system(GIS)technology is used to integrate multi-camera video data,and construct the integrated information of video moving objects and geographical scene by video-GIS fusion.After that,it lays a foundation for the following multi-camera video moving objects’ trajectories clustering,objects retrieval and video synopsis.2.Through the analysis of geographical environment and multi-camera video moving objects’ trajectories in geographical scene,we can optimize the single camera video moving objects’ trajectories clustering technology and explore it to multi-camera environment.Then,the overall movement trend of moving objects in the geographical scene could be mined.3.By retrieval of moving objects under specific conditions(spatial and spatiotemporal constraints),the objects of video synopsis could be accurately located,which can not only shorten the number of video frames associated with the retrieval object,but also provide support for the research on the condensed length of video synopsis.4.Through analyzing the topological relationship between multi-camera location and view in geographic scene,cameras are grouped(videos are divided by along with cameras).After that,we can explore the method of multi-camera video synopsis based on the cross camera trajectory association of video frames and moving objects.